PH and EBM Flashcards
PICO
patient/ population
Intervention/ treatment
Comparison/ alternative treatment
Other relevant clinical info
types of observational study
advantages and disadvantages
observational:
- ecological
- cross sectional
- case-control
- cohort study
advantages:
- ethics (can’t force a group to smoke)
- can use very large groups
disadvantages:
- biases
- confounding (links an exposure with an outcome)
- reverse causality
cohort study
DONT HAVE ANY DISEASE AT START - see who develops disease with risk factors
- exposure to defined factors measured at baseline
- any new incidence of disease
- high and low exposure individuals compared
- calculate RISK RATIO
- prospective or retrospective
advantages:
- reduce reverse causality
- reduces selection bias
- allows testing of multiple outcome
- better confounder control
disadvantages:
- retrospective: recall and interviewer bias, reverse causality
- prospective: LONG, LOSS TO FOLLOW UP bias
- inefficient for rare diseases
cross-sectional study
SNAPSHOT OF PREVALENCE
- shows prevalence of disease in population at snap shot moment,
- good for measuring true burden of disease
- measure risk factors,
- can’t measure incidence, susceptible to reverse causality
case-control study
WHAT IS CAUSE OF THE OUTCOME
- recruit available cases and a comparable control group
- sample determined by outcome
- RETROSPECTIVELY assess exposure to potential risk factors - compare case and control
- if exposure more common in cases, risk factor associated with increased disease
- presented as ODDS RATIO
advantages:
- only study comparing groups defined by outcome
- good for RARE CONDITIONS
- can test for multiple exposures
disadvantages:
- reverse causality
- selection bias: when selecting control group, choose suitable population, not one which would have higher/lower associations to exposure
- measurement bias:
recall bias- cases more likely to recall exposure as they understand disease
interviewer bias - can cause cases to recall more exposure
systematic review
- all relevant evidence on a given clinical question
- minimise biases and random errors
- highest quality of evidence
- much cheaper and quicker than RCTs
disadvantages:
- only as good as the studies they’re done on
- reporting biases:
- -> publication bias - statistically significant are more likely published
- -> time lag - big studies published quicker
- -> language - English get published quicker
- -> multiple publication - big studies may be published in multiple places
p value
the probability of the observed results occurring just by chance, if the null hypothesis was true
very low p-value indicates strong evidence against the null hypothesis (differs in outcome to control)
risk ratio vs odds ratio
risk ratio: out of total number of people in the study
odds ratio: out of unaffected patients
sensitivity vs specificity
sensitivity:
true positive rate - probability of a positive test in people with disease (proportion of all those with the condition)
specificity:
true negative rate - prob of negative test result in people without disease
primary prevention definition
prevent the onset of disease
secondary prevention
early identification and treatment of disease
tertiary prevention
rehabilitate people with established disease
prevention paradox
large number of small risk cases may get disease, but population interventions may provide little to individuals
R number
effective reproduction rate - number coming from one case
R0
basic reproduction number R0 =
probability of effective contact x number of contacts x duration of infectiousness
better for planning for infectious diseases
(in a wholly susceptible population- scenario with no immunity)
cost effectiveness
cost per clinical effect
cost utility
cost per QALY
SMR ratio calculation
SMR =
example q:
condition present:
test +ve: 60
test -ve: 57
total: 117
condition absent:
test +ve: 1
test -ve: 400
total: 401
total +ve: 61
total -ve: 457
total people:
518
calculate sensitivity, specificity, positive predictive value, negative predictive value
sensitivity: 60/117
specificity: 400/401
positive predictive value: 60/61
negative predictive value: 400/457
positive predictive value
probability of having disease if you test positive
negative predictive value
probability of not having disease if you test negative
inverse equity hypothesis
new health interventions are initially adopted by wealthy/ educated, initially increasing inequalities as poorest/ less educated lag behind on uptake of the new health intervention
example: traffic light labelling of foods sold pre-packaged
which mechanism is bias reduced by with using big groups (e.g. school) instead of individuals
contamination
what aspect of trial quality is always feasible when comparing treatments in an RCT
allocation concealment
not: blinding (on either side), follow up
best available evidence in order
- systematic review
- RCTs
- cohort studies
- case-controlled studies
- background info
ecological study
WHAT IS THE EXPOSURE CAUSING
- average exposure plotted against rate of outcome
- association btwn them
- environmental or social exposures at population levels
- disadvantages: ecological fallacy
difficult to control confounding
dependent on previously collected data
what is ecological fallacy?
disadvantage of ecological studies: the assumption that average characteristics apply to individual
e.g. smokers get cancer
RCTs
- interventional study
- two arms one group exposed
- participants consent
advantages:
- evidence of causality
- best confounder control
- allocation concealment reduces selection bias
- blinding reduces measurement bias
disadvantage:
- sample size needs to be large enough
- selection bias
- performance bias
- detection bias
- attrition bias (unequal loss of participants)
meta analysis
statistical analysis - combing results of independent studies
- similar intervention, similar outcome, similar populations
fixed (common) effect:
- assumes true effect is the same in each study.
- the only variation in estimates is sampling error. (assumes all studies are trying to show the same thing- homogeneity)
- less weight given to small samples
random effects:
- estimates mean effect (assumed that the true study effects vary- not showing same effect - heterogeneity).
- info from small studies matters more
heterogeneity:
- suggests treatment effect is context dependent (all studies not showing same effect)
funnel plots asymmetry:
- publication/reporting bias, poorer quality studies –> extreme treatment effects
what methods are used for qualitative studies
for helping understand why something happened
- observations
- interviews
- focus groups
- documents
- oral history
what data analysis is used for qualitative studies?
- familiarisation with data
- coding (repeated ideas)
- searching for themes
- reviewing themes
- defining and naming themes
- writing up
performance bias
affects RCTs
systematic differences in the care provided to members of different study groups other than the intervention
detection bias
affects RCTs
systematic differences btwn groups in how outcomes are determined
attrition bias
affects RCTs
if the numbers lost to follow up are not the same in each group
intention to treat analysis
done to avoid the effects of crossover and dropout
comparison of the treatment groups that includes all patients as originally allocated after randomization
per protocol analysis
instead of intention to treat analysis
comparison of treatment groups that includes only those patients who completed the treatment originally allocated
reporting biases
- publication bias- statistically significant are more likely to be published
- time lag - big studies published quicker
- language - English studies quicker
- multiple publication - in bigger studies
forest plots
fixed or random model
pulled odds ration
with 95% CI tails
diamond: overall odds ratio in the middle with overall 95% CI on the ends
funnel plots
to show whether there is publication bias
- used in SRs and meta-analysis
- asymmetric (if part not filled in, they haven’t been published –> publication bias)
prevalence calculation
total patients with disease in population
individuals w disease
/
total population at risk
incidence
new cases in given period
/
population at risk initially disease free
incidence rate
(# new cases of disease)
/
(population at risk) x time interval)
risk
patients w disease
/
population
risk ratio
risk of outcome occurrence in exposed
/
risk of outcome occurrence in unexposed
likelyhood of disease with exposure compared to without exposure- strength of association not causality
RR> 1 exposure predisposes outcome
RR<1 exposure protects against outcome
risk difference
risk of outcome in exposed - risk of outcome in unexposed
number needed to treat
1/ risk difference
number of patients that would need to be treated to prevent one case of the disease
odds
patients with disease
/
patients without disease
odds ratio
odds of having disease in exposed
/
odds of having disease in unexposed
null hypothesis
- no difference btwn case population and control population
- study aims to disprove null hypothesis
p-value
- the probability that differences in observed data would have occurred by chance
- small p-value (p<0.05) = greater evidence against null hypothesis
incremental cost-effectiveness ratio
= C/E
= (Ct - Cc)/ (Et - Ec)
TOTAL COST OVER QALY
C= cost E= effectiveness t = treatment c = comparator
compares treatment to comparator (next best)
Quality Adjusted Life Year (QALY)
= (time spent in healthy state) x (quality of life weight)
standard mortality ratio
= (# observed deaths / # expected deaths) x 100
- ratio of observed deaths to expected deaths
- SMR 100 = study pop has same number of deaths as standard pop
- SMR > 100 more than expected # of deaths
statistical power
probability of correctly rejecting null hypothesis when in truth the treatment has an effect
net monetary benefits
= (E x lamda) - C
E= cost lamda = willingness to pa for QALY C = cost
95% CI
range of value within which we are 95% confident that the true population value lies
+ve likelihood ratio
probability of a positive test in people with the disease/ probability of a positive test in people without the disease
= sensitivity/ (1-specificity)
used with layers monogram
accuracy
(true positives + true negatives)
/
all results
SpPin
when a test has a high specificity a positive result rules IN the target disorder
SnNout
sensitivity - when a test has a high sensitivity a negative result rules out the target disorder
spectrum bias in diagnostic testing
the types of patients recruited to the study
work up bias in diagnostic testing
do all patients get both diagnostic and gold standard tests
improving public health by health protection
infectious disease - childhood vaccination - immunisation environmental hazards emergency response to infectious disease outbreaks
improving public health by health promotion
- develop primary promotion programme
- health inequalities
- behaviour change
improving public health by health services
- secondary prevention programmes e.g. screening
- healthcare quality
- health policy
proportionate universalism
- what is it the solution to and what is it?
solution to prevention paradox
providing service universally but with increased intensity on disadvantaged
vaccination programme types
- universal rolling - give vaccine continuously to everyone of certain age when they reach it
- universal catch-up - give vaccine to everyone in pop within a certain age range with a fixed time period
- targeted - aimed at people who are high risk
types of vaccine
live vaccine:
- MMR, BCG, rotavirus, polio, influenza
- strong immune response
- can be administered via mucosal route
- mild infection after
- must be maintained alive (cold chain)
conjugate vaccine:
- HiB, men C
- development of immunity to non-protein material
- smaller response
toxoid:
- diphtheria, tetnus
- no risk of infection
subunit
- pertussis, MenB
health inequality vs health inequity
health inequality: differences in health
health inequity: unjust difference
harms of screening
- false negatives/ false reassurance
- over detection - false anxiety
- can increase incidence (picking up more less serious cases)
- length time bias - leaves out poor prognosis (cause treated quickly)
healthy screened effect
people who come for screening tend to be healthier than those who don’t
